【摘要】 针对高光谱遥感图像去噪问题,提出了基于多任务非负字典学习的高光谱图像去噪算法。该算法基本思想是通过各波段图像稀疏表示系数矩阵一致性约束,利用高光谱遥感图像各波段之间的强相关性提高去噪性能。首先,建立高光谱遥感图像的多任务非负字典学习模型。然后构造迭代格式求解该模型得到各波段图像的冗余字典和共同的稀疏表示系数矩阵。最后利用各波段冗余字典和共同的系数矩阵复原图像。相比较现有先进的算法,由于充分利用了高光谱图像各波段的强相关性这一内在特征,使得文中算法能够很好地保持各波段图像的空间细节和光谱信息。实验结果验证了该算法的有效性。更多还原

【Abstract】 For the hyperspectral remote sensing image denoising problem, the image denoising algorithm based on multitask nonnegative dictionary learning is proposed. The basic idea of the proposed algorithm is to improve image denoising effects by using the high correlation between each band image by sharing a common coefficient matrix. Firstly, the multitask nonnegative dictionary learning model is constructed. Then this model is solved by the iterative method to obtain the redundant dictionary of each band and the common sparse representation coefficient matrix. Finally, clean images can be restored by each band redundant dictionary and the common coefficient matrix. Compared with state-of-the-art methods, the proposed algorithm can well preserve the spatial detail and spectral information of each band image for taking advantage of the high correlation between each band. Experimental results validate the effectiveness of the proposed algorithm.更多还原